Related Experiment Video
Updated: May 12, 2026
A Novel Digital Platform for a Monitored Home-based Cardiac Rehabilitation Program
Published on: April 19, 2019
Predicting Home Exercise Adherence after Ischemic Stroke: Development and Validation of a Web-Based Nomogram
1Department of Clinical Nursing Education, The Second Affiliated Hospital of Harbin Medical University, Harbin, China.
Background:
Poor adherence to home-based functional exercises can substantially hinder recovery after ischemic stroke. This study aimed to develop and validate a web-based predictive nomogram to identify patients with ischemic stroke who are at risk of poor adherence to home-based functional exercises.
Methods:
We conducted a cross-sectional study of 536 patients with ischemic stroke and limb dysfunction at a tertiary hospital in China. Latent profile analysis (LPA) was used to classify adherence patterns based on the Exercise Adherence Questionnaire. The least absolute shrinkage and selection operator (LASSO) regression was applied to select predictors from 35 candidate variables, and multivariable logistic regression was then used to build the prediction model. Model performance was internally validated using 1000 bootstrap resamples and assessed in terms of discrimination, calibration, and clinical utility. A web-based nomogram was developed for clinical use.
Results:
The sample included 254 males (47.4%) and 282 females (52.6%); 61.0% were aged ≥ 60 years. LPA identified three adherence profiles: low (18.1%), moderate (42.2%), and high (39.7%). The optimal cutoff score for distinguishing good from poor adherence was 36.5 points. Five independent predictors were retained: marital status (never married: odds ratio [OR] = 0.03, 95% confidence interval [CI]: 0.01-0.11), monthly income > 5000 RMB (OR = 0.31, 95% CI: 0.14-0.67), spouse as primary caregiver (OR = 0.23, 95% CI: 0.10-0.53), knowledge level (OR = 0.92, 95% CI: 0.87-0.98), and exercise motivation (OR = 0.87, 95% CI: 0.81-0.93). Internal validation using 1000 bootstrap resamples showed good discrimination (apparent C-statistic = 0.858; optimism-corrected C-statistic = 0.848), good calibration (optimism-corrected calibration slope = 0.939; calibration intercept = 0.009), and a positive net benefit across threshold probabilities ranging from 0 to 0.80. After uniform shrinkage, model performance remained stable.
Conclusions:
We developed and internally validated a prediction model combining LPA with LASSO-logistic regression. The web-based nomogram may facilitate early identification of patients at risk of poor adherence to home-based functional exercises and support targeted interventions to improve rehabilitation outcomes.
